Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS20040247054 A1
Publication typeApplication
Application numberUS 10/814,808
Publication dateDec 9, 2004
Filing dateApr 1, 2004
Priority dateApr 1, 2003
Also published asDE602004016587D1, EP1465338A2, EP1465338A3, EP1465338B1, US7336734
Publication number10814808, 814808, US 2004/0247054 A1, US 2004/247054 A1, US 20040247054 A1, US 20040247054A1, US 2004247054 A1, US 2004247054A1, US-A1-20040247054, US-A1-2004247054, US2004/0247054A1, US2004/247054A1, US20040247054 A1, US20040247054A1, US2004247054 A1, US2004247054A1
InventorsAnne Ferreol, Laurent Albera, Pascal Chevalier
Original AssigneeAnne Ferreol, Laurent Albera, Pascal Chevalier
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Method and device for the fourth-order, blind identification of an under-determined mixture of sources
US 20040247054 A1
Abstract
A method for the fourth-order, blind identification of at least two sources in a system comprising a number of sources P and a number N of reception sensors receiving the observations, said sources having different tri-spectra. The method comprises at least the following steps: a step for the fourth-order whitening of the observations received on the reception sensors in order to orthonormalize the direction vectors of the sources in the matrices of quadricovariance of the observations used; a step for the joint diagonalizing of several whitened matrices of quadricovariance in order to identify the spatial signatures of the sources. Application to a communication network.
Images(7)
Previous page
Next page
Claims(20)
1. A method of fourth-order, blind identification of two sources in a system including a number of sources P and a number N of reception sensors receiving the observations, the sources having different tri-spectra, comprising the following steps:
a) fourth-order whitening of the observations received on the reception sensors in order to orthonormalize the direction vectors of the sources in the matrices of quadricovariance of the observations used,
b) joint diagonalizing of several whitened matrices of quadricovariance to identify the spatial signatures of the sources.
2. The method according to claim 1, wherein the observations used correspond to the time-domain averaged matrices of quadricovariance defined by:
Q x ( τ 1 , τ 2 , τ 3 ) = p = 1 P c p ( τ 1 , τ 2 , τ 3 ) ( a p a p * ) ( a p a p * ) H (4a) = A Q C s ( τ 1 , τ 2 , τ 3 ) A Q H (4b)
where AQ is a matrix with a dimension (N2×P) defined by AQ=[(α1{circle over (×)}α1*), . . . , (αp{circle over (×)}αp*)], Cs123) is a diagonal matrix with a dimension (P×P) defined by Cs123)=diag[cl123), . . . , cp123)] and cp123) is defined by:
c p123)=<Cum(s p(t), s p(t−τ 1)*, s p(t−τ 2)*, s p(t−τ 3))>  (5)
3. The method according to claim 2, comprising the following steps:
Step 1: estimating, through Q;{circumflex over ( )}x, of the matrix Qx, from the L observations x(ITe) using a non-skewed and asymptotically consistent estimator.
Step 2: eigen-element decomposition of Q;{circumflex over ( )}x, the estimation of the number of sources P and the limiting of the eigen-element decomposition to the P main components: Q;{circumflex over ( )}x≈E;{circumflex over ( )}xΛ;{circumflex over ( )}xE;{circumflex over ( )}x H, where Λ;{circumflex over ( )}x is the diagonal matrix containing the P eigenvalues with the highest modulus and E;{circumflex over ( )}x is the matrix containing the associated eigenvectors.
Step 3: building of the whitening matrix: T;{circumflex over ( )}=(Λ;{circumflex over ( )}x)−1/2E;{circumflex over ( )}x H.
Step 4: selecting K triplets of delays (τ1 k2 k3 k) where |τ1 k|+|τ2 k|+|τ3 k≠0.
Step 5: estimating, through Q;{circumflex over ( )}x1 k2 k3 k), of the K matrices Qx1 k2 k3 k).
Step 6: computing of the matrices T;{circumflex over ( )} Q;{circumflex over ( )}x1 k2 k3 k) T;{circumflex over ( )}H and the estimation, by U,{circumflex over ( )}sol, of the unitary matrix Usol by the joint diagonalizing of the K matrices T;{circumflex over ( )} Q;{circumflex over ( )}x1 k2 k3 k) T;{circumflex over ( )}H
Step 7: computing T;{circumflex over ( )}#U;{circumflex over ( )}sol=[b;{circumflex over ( )}1 . . . b;{circumflex over ( )}P] and the building of the matrices B;{circumflex over ( )}1 sized (N×N).
Step 8: estimating, through α;{circumflex over ( )}P, of the signatures aq (1≦q≦P) of the P sources in applying a decomposition into elements on each matrix B;{circumflex over ( )}1.
4. The method according to claim 1, comprising evaluating quality of the identification of the associated direction vector in using a criterion:
D(A, Â)=(α1, α2, . . . , αP)   (16)
where
α p = min 1 i P [ d ( a p , a ^ i ) ] ( 17 )
and where d(u,v) is the pseudo-distance between the vectors u and v, such that:
d ( u , v ) = 1 - u H v 2 ( u H u ) ( v H v ) ( 18 )
5. The method according to claim 1, a fourth-order cyclical after the step a) of fourth-order whitening.
6. The method according to claim 5, wherein the identification step is performed in using fourth-order statistics.
7. The method according to claim 1 wherein the number of sources P is greater than or equal to the number of sensors.
8. The method according to claim 1, comprising goniometry using the identified signature of the sources.
9. The method according to claim 1, comprising spatial filtering after the identified signature of the sources.
10. The use of the method according to claim 1, for use in a communications network.
11. The method according to claim 2, comprising evaluating quality of the identification of the associated direction vector in using a criterion
D ( A , A ^ ) = ( α 1 , α 2 , , α P ) where α p = min 1 i P [ d ( a p , a ^ i ) ]
and where d(u,v) is the pseudo-distance between the vectors u and v, such that:
d ( u , v ) = 1 - u H v 2 ( u H u ) ( v H v )
12. The method according to claim 3, comprising evaluating quality of the identification of the associated direction vector in using a criterion
D ( A , A ^ ) = ( α 1 , α 2 , , α P ) where α p = min 1 i P [ d ( a p , a ^ i ) ]
and where d(u,v) is the pseudo-distance between the vectors u and v, such that:
d ( u , v ) = 1 - u H v 2 ( u H u ) ( v H v )
13. The method according to claim 2, a fourth-order cyclical after the step a) of fourth-order whitening.
14. The method according to claim 2, wherein the identification step is performed in using fourth-order statistics.
15. The method according to claim 2, wherein the number of sources P is greater than or equal to the number of sensors.
16. The method according to claim 2, comprising goniometry using the identified signature of the sources.
17. The method according to claim 3, a fourth-order cyclical after the step a) of fourth-order whitening.
18. The method according to claim 3, wherein the identification step is performed in using fourth-order statistics.
19. The method according to claim 3, wherein the number of sources P is greater than or equal to the number of sensors.
20. The method according to claim 3, comprising goniometry using the identified signature of the sources.
Description
BACKGROUND OF THE INVENTION

[0001] 1. Field of the Invention

[0002] The invention relates especially to a method for the learned or blind identification of a number of sources P that is potentially greater than or equal to the number N of sensors of the reception antenna.

[0003] It can be used for example in the context of narrow-band multiple transmission.

[0004] It is used for example in a communications network.

[0005] It can be applied especially in the field of radio communications, space telecommunications or passive listening to these links in frequencies ranging for example from VLF to EHF.

[0006]FIG. 1 is a schematic drawing exemplifying an array of several reception sensors or receivers, each sensor receiving signals from one or more radio communications transmitters from different directions of arrival

[0007] Each sensor receives signals from a source with a phase and amplitude that are dependent on the angle of incidence of the source and the position of the sensor. FIG. 2 is a drawing exemplifying the parametrization of the direction of a source. This direction is parametrized by two angles corresponding to the azimuth angle θ and the elevation angle Δ.

[0008] 2. Description of the Prior Art

[0009] The past 15 years or so have seen the development of many techniques for the blind identification of signatures or source direction vectors, assumed to be statistically independent. These techniques have been developed in assuming a number of sources P smaller than or equal to the number of sensors N. These techniques have been described in the references [1][3][7] cited at the end of the description. However, for many practical applications such as HF radio communications, the number of sources from which signals are received by the sensors is increasing especially with the bandwidth of the receivers, and the number of sources P can therefore be greater than the number of sensors N. The mixtures associated with the sources are then said to be under-determined.

[0010] A certain number of methods for the blind identification of under-determined mixtures of narrow-band sources for networks have been developed very recently and are described in the references [2] [7-8] and [10]. The methods proposed in the references [2] and [7-8] make use of information contained in the fourth-order (FO) statistics of the signals received at the sensors while the method proposed in the reference [10] make use of the information contained in one of the characteristic functions of the signals received. However, these methods have severe limitations in terms of the prospects of their operational implementation. Indeed, the method described in the reference [2] is very difficult to implement and does not provide for the identification of the sources having the same kurtosis values (standardized fourth-order cumulant). The methods described in the references [7-8] assume that the sources are non-circular. These methods give unreliable results in practice. Finally, the reference method [10] has been developed solely for mixtures of sources with real (non-complex) values.

[0011] The object of the present invention relates especially to a new method for the blind identification of an under-determined mixture of narrow-band sources for communications networks. The method can be used especial y to identify up to N2−N+1 sources from N identical sensors and up to N2 sources with N different sensors, in assuming only that the sources have different tri-spectra and non-zero, same-sign kurtosis values (this hypothesis is practically always verified in the context of radio communications).

SUMMARY OF THE INVENTION

[0012] The invention relates to a method for the fourth-order, blind identification of at least two sources in a system comprising a number of sources P and a number N of reception sensors receiving the observations, said sources having different tri-spectra, wherein the method comprises at least the following steps:

[0013] a step for the fourth-order whitening of the observations received on the reception sensors in order to orthonormalize the direction vectors of the sources in the matrices of quadricovariance of the observations used,

[0014] a step for the joint diagonalizing of several whitened matrices of quadricovariance in order to identify the spatial signatures of the sources.

[0015] The number of sources P is for example greater than the number of sensors N.

[0016] The method can be used in a communication network.

[0017] The method according to the invention has especially the following advantages:

[0018] it enables the identification of a number of sources greater than the number of sensors:

[0019] for identical sensors: N2−N+1 sources having different tri-spectra and non-zero and same-sign kurtosis values,

[0020] for different sensors (arrays with polarization diversity and/or pattern diversity and/or coupling etc) N2 sources having different tri-spectra and non-zero, same-sign kurtosis values,

[0021] the method is robust with respect to Gaussian noise, even spatially correlated Gaussian noise,

[0022] it enables the goniometry of each source identified, using a wavefront model attached to the signature, with a resolution potentially higher than that of existing methods,

[0023] it enables the identification of I (N2−N+1) cyclostationary sources if the sensors are identical and I×N2 cyclostationary sources if the sensors are different: with polarization diversity and/or pattern diversity and/or coupling, where I is the number of cyclical frequencies processed,

[0024] using a performance criterion, it enables the quantitative evaluation of the quality of estimation of the direction vector of each source and a quantitative comparison of two methods for the identification of a given source,

[0025] using a step for the selection of the cyclical frequencies, it enables the processing of a number of sources greater than the number of sources processed by the basic method.

BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Other features and advantages of the invention shall appear more clearly from the following description along with the appended figures, of which:

[0027]FIG. 1 shows an example of a communication network,

[0028]FIG. 2 shows parameters of a source,

[0029]FIG. 3 is a functional diagram of the method according to the invention,

[0030]FIG. 4 shows an example of spatial filtering,

[0031]FIGS. 5 and 6 show examples of variations of the performance criterion as a function of the number of samples observed, comparing the performance of the method with two prior art methods.

[0032] FIGS. 7 to 9 show three alternative embodiments of the method described in FIG. 3 implementing a selection of the cyclical frequencies.

MORE DETAILED DESCRIPTION

[0033] For a clear understanding of the object of the invention, the following example is given by way of an illustration that in no way restricts the scope of the invention for a radio communications network in a multiple-transmission, narrow-band context, with sources having different tri-spectra (of cumulants).

[0034] Each sensor of the network, formed by N receivers, receives a mixture of P narrow-band (NB) sources which are assumed to be statistically independent. On this assumption, the vector of the complex envelopes of the signals at output of the sensors is written as follows: x ( t ) = p = 1 P s p ( t ) a p + b ( t ) = A s ( t ) + b ( t ) ( 1 )

[0035] where sp(t) is the signal of the pth source as well as the pth component of the vector s(t), b(t) is the assumed Gaussian noise vector with any covariance, αp is the signature or the direction vector of the pth source and A (N×P) is the matrix of the vectors αp (direction vectors of the sources).

[0036] It is an object of the invention especially to identify the direction vectors αp of each of the sources when, especially, the number of sources P is potentially greater than the number of sensors N.

[0037] From this identification, it is then possible to apply techniques for the extraction of the sources by the spatial filtering of the observations. The blind extraction is aimed especially at restoring the information signals conveyed by the sources in not exploiting any a priori information (during normal operation) on these sources.

[0038] Fourth-Order Statistics

[0039] The method according to the invention makes use of the fourth-order statistics of the observations corresponding to the time-domain averages Qx123)=<Qx123)(t)>, on an infinite horizon of observation, of certain matrices of quadricovariance, Qx123)(t), , sized (N2×N2). The elements, Qx123)[i, j, k, l](t) of these matrices are, for example, defined by the relationship:

Q x123)[i,j,k,l][t]=Cum(x i(t),x j(t−τ 1)*,x k(t−τ 2)*, x l(t−τ 3))   (2)

[0040] where * is the conjugate complex symbol, xi(t) is the ith component of the vector x(t), <.> is the operation of time-domain averaging on an infinite horizon of observation and (τ123) is a triplet of delays. Assuming that Qx123)[i, j, k, l] is the element [N(i−1)+j, N(k−1)+l] of the matrix Qx123), assuming that the noise is Gaussian and using the expression (1) in the expression (2), the matrix of quadricovariance Qx123) is written as follows:

Q x123)=(A{circle over (×)}A*)Q s123)(A{circle over (×)}A*)H   (3)

[0041] where Qs123) is the averaged matrix of quadricovariance of s(t) with a dimension (P2×P2), A=[α1 . . . αP], {circle over (×)} is the Kronecker product and H designates the transpose and conjugate. Assuming statistically independent sources, the matrix Qs123) is formed by at least P4−P zeros and the expression (3) is simplified as follows: Q x ( τ 1 , τ 2 , τ 3 ) = p = 1 P c p ( τ 1 , τ 2 , τ 3 ) ( a p a p * ) ( a p a p * ) H (4a) = A Q C s ( τ 1 , τ 2 , τ 3 ) A Q H (4b)

[0042] where AQ is a matrix sized (N2×P) defined by AQ=[(α1{circle over (×)}α1*), . . . , (αp{circle over (×)}αp*)], Cs123) is a diagonal matrix sized (P×P) defined by Cs123)=diag[c1123), . . . , cp123)] and where cp123) is defined by:

c p123)=<Cum(s p(t), s p(t−τ 1)*, s p(t−τ 2)*, s p(t−τ 3))>  (5)

[0043] The expression (4b) has an algebraic structure similar to that of the correlation matrix of the observations used in the algorithm SOBI described in the reference [1]. The notation used here below will be Qx=Qx(0, 0, 0), cp=cp(0, 0, 0), Cs=Cs(0, 0, 0) in order to deduce the relationship (4b) therefrom:

Qx=AQCsAQ H   (6)

[0044] It is assumed here below that the number of sources P is such that P≦N2, that the matrix AQ is of full rank, that the averaged cumulants cp, 1≦p≦P, are non-zero (non-Gaussian sources) and same-sign cumulants and that, for any pair (i, j) of sources, there is at least one triplet of delays (τ123) such that |τ1|+|τ2|+|τ3|=0 and

c i123)/|c i |≠c j123)/|c j|  (7)

[0045] Fourth-Order Whitening Step

[0046] The first step of the method according to the invention, called FOBIUM, consists of the orthonormalization, in the matrix of quadricovariance Qx of the expression (6), of the columns of the matrix AQ, considered to be virtual direction vectors of the sources for the array of sensors considered. To this end, the method considers the eigen-element decomposition of the P rank hermitian matrix Qx given by

Qx=ExΛxEx H   (8)

[0047] where Ax is the real longitudinal diagonal, with a dimension (P×P), of the P non-zero eigenvalues of Qx, and Ex is the matrix sized (N2×P) of the associated, orthonormal eigenvectors. For a full-rank matrix AQ, it can be shown that there is equivalence between assuming that the kurtosis values of the sources have a same sign ε (ε=±1) and assuming that the eigenvalues of Λx also have a same sign ε. In this context, it is possible to build the following whitening matrix T sized (P×N2):

T=(Λ x)−1/2 E x H   (9)

[0048] The whitening matrix sized (P×N2) is defined from the square root of the real diagonal matrix sized (P×P) of the P non-zero eigenvalues of the matrix of quadricovariance and of the transpose of the matrix of the associated eigenvectors with a dimension (P×N2) where (Λx)−1/2 is the inverse of the square root of Λx. From the expressions (6) and (8), it is deduced that:

εTQ x T H =TA QC s)A Q H T H =I P   (10)

[0049] where IP is the identity matrix with a dimension (P×P) and where εCs=diag[|c1|, . . . , |cp|]. This expression shows that the matrix TAQ(εCs)1/2 with a dimension (P×P) is a unitary matrix U. It is deduced from this that:

TA Q =UC s)−1/2   (11)

[0050] Fourth-Order Identification Step

[0051] From the expressions (4b) and (11), it is deduced that:

W123)=TQ x123)T H =UC x)−1/2 C s123)(εC s)−1/2 U H   (12)

[0052] where W(τ123) is the matrix of quadricovariance whitened at the fourth order by the matrix Qx. This expression shows that the unitary matrix U diagonalizes the matrices T Qx123) TH and that the associated eigenvalues are the diagonal terms of the diagonal matrix (εCs)−1/2 Cs123) (εCs)−1/2. For a given triplet of delays (τ123), the matrix U is unique, give or take one permutation and one unit diagonal matrix, when the elements of the matrix (εCs)−1/2 Cs123) (εCs)−1/2 ) all are different. If not, the method uses a set of K triplets (τ1 k2 k3 k), 1≦k≦K, defined as follows: for all pairs of sources (i, j), there is at least one triplet (τ1 k2 k3 k), such that the condition of the equation (7) is verified. In these conditions, the unitary matrix U is the only matrix Usol which, to the nearest permutation and to the nearest unit diagonal matrix, jointly diagonalizes the K matrices T×Qx1 k2 k3 k)×TH. Consequently, the matrix Usol, which resolves the above problem, is written as a function of the unitary matrix U as follows:

Usol=UΛΠ  (13)

[0053] where Λ and Π are respectively the unit diagonal matrix and the permutation matrix referred to here above. From the equations (11) and (13), it is possible to deduce the matrix AQ to the nearest unit diagonal matrix and to the nearest permutation, which is expressed by:

T # U sol =[b 1 . . . b P ]=E xΛx 1/2 U sol =A QC s)1/2Λ  (14)

[0054] where T# is the pseudo-inverse of the matrix T. Each column, bl (1≦l≦P), of the matrix T# Usol corresponds to one of the vectors μq |cq|1/2 q{circle over (×)}αq*), 1≦q≦P, where μq is a complex scalar value such that |μq|=1. Consequently, in converting each column bl of the matrix T# Usol into a matrix Bl with a dimension (N×N such that Bl[i, j]=bl((i−1)N+j) (1≦i, j≦N), it is deduced therefrom that:

B lq |c q|1/2αqαq H pour (1≦l, q≦P)   (15)

[0055] The matrix Bl is built from the vector bl and depends on a complex scalar value, the square root of the cumulant and the direction vector of the qth source and of its conjugate.

[0056] In this context, the direction vector αq of the qth source is associated with the eigenvector of Bl associated with the highest eigenvalue.

SUMMARY OF THE PRINCIPLE OF THE INVENTION

[0057] In brief, the different steps of the method according to the invention include at least the following steps: for L vector observations received in the course of the time: x(lTe) (1≦l ≦L), where Te is the sampling period.

[0058] Estimation

[0059] Step 1: The estimation, through Q;{circumflex over ( )}x, of the matrix of quadricovariance Qx, from the L observations x(lTe), using a non-skewed and asymptotically consistent estimator. Depending on the nature of the sources, the estimator is adapted as follows:

[0060] Stationary and centered case: empirical estimator used in the reference [3].

[0061] Cyclostationary and centered case: estimator implemented in the reference [10].

[0062] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0063] Whitening

[0064] Step 2: The eigen-element decomposition of the estimated matrix of quadricovariance Q;{circumflex over ( )}x, estimating the number of sources P and restricting this eigenvalue decomposition to the P main components: Q;{circumflex over ( )}x≈E;{circumflex over ( )}x Λ;{circumflex over ( )}x E;{circumflex over ( )}x H, where A;{circumflex over ( )}x is the diagonal matrix containing the P highest modulus eigenvalues and E;{circumflex over ( )}x is the matrix containing the associated eigenvectors.

[0065] Step 3: The building of the whitening matrix: T;{circumflex over ( )}=(Λ;{circumflex over ( )}x)−1/2 E;{circumflex over ( )}x H.

[0066] Selection of the Triplets

[0067] Step 4: The selection of K triplets of delays (τ1 k2 k3 k) where |τ1 k|+|τ2 k|+|τ3 k|≠0.

[0068] Estimation

[0069] Step 5: The estimation, through Q;{circumflex over ( )}x1 k2 k3 k), of K matrices of quadricovariance Qx1 k2 k3 k). As in the step 1, this estimation depends especially on the assumptions made on the observations:

[0070] Stationary and centered case: empirical estimator used in the reference [3].

[0071] Cyclostationary and centered case: estimator implemented in the reference [10].

[0072] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0073] Identification

[0074] Step 6: The computation of the matrices T;{circumflex over ( )} Q;{circumflex over ( )}x1 k2 k3 k) T;{circumflex over ( )}H and the estimation, by U;{circumflex over ( )}sol, of the unitary matrix Usol through the joint diagonalizing of the K matrices T;{circumflex over ( )} Q;{circumflex over ( )}x1 k2 k3 k) T;{circumflex over ( )}H.

[0075] Step 7: The computation of T{circumflex over ( )} #U;{circumflex over ( )}sol=[b;{circumflex over ( )}l . . . b;{circumflex over ( )}P] and the building of the matrices B;{circumflex over ( )}l sized (N×N).

[0076] Step 8: The estimation, by α;{circumflex over ( )}P, of the signatures αq (1≦q≦P) of the P sources in carrying out a decomposition into elements on each matrix B;{circumflex over ( )}l.

[0077] Applications

[0078] At the end of the step 8, the method has identified the direction vectors of P non-Gaussian sources having different tri-spectra with same-sign kurtosis values. p<N2 and P may reach N2−N+1 or N2 depending on the type of sensors used.

[0079] Using this information, the method may implement a method of goniometry or a spatial filtering of antennas.

[0080] A method of goniometry can be used to determine the direction of arrival of the sources and more precisely the azimuth angle θm for 1D goniometry and azimuth and elevation angles (θm, Δm) for 2D goniometry.

[0081]FIG. 4 represents a spatial filtering of antennas for spatial filtering structures. It enables especially the optimizing of reception from one or all the sources present by the spatial filtering of the observations. When several sources are of interest to the receiver, we speak of source separation techniques. When no a priori information on the sources is exploited, we speak of blind techniques.

[0082] Verification of the Quality of the Estimates

[0083] According to one alternative embodiment, the method comprises a step of qualitative evaluation, for each source, of the quality of identification of the associated direction vector.

[0084] This new criterion enables the intrinsic comparison of two methods of identification for the restitution of the signature of a particular source. This criterion, for the identification problem, is an extension of the one proposed in [5] for extraction. It is defined by the P-uplet

D(A, Â)=(α1, α2, . . . , αP)   (16)

[0085] where α p = min 1 i P [ d ( a p , a ^ i ) ] ( 17 )

[0086] and where d(u,v) is the pseudo-distance between the vectors u and v, such that: d ( u , v ) = 1 - u H v 2 ( u H u ) ( v H v ) ( 18 )

[0087] In the simulations of FIGS. 5 and 6, there are P=6 statistically independent sources received on a circular array of N=3 sensors having a radius r such that r/λ=0.55 (λ: wavelength). The six sources are non-filtered QPSK sources having a signal-to-noise ratio of 20 dB with a symbol period T=4Te, where Te is the sampling period.

[0088] The incidence values of the sources are such that θ1=2.16°, θ2=25.2°, θ3=50°, θ4=272.16°, θ5=315.36°, θ6=336.96° and the associated carrier frequencies verify Δf1 Te=0, Δf2 Te=1/2, Δf3 Te=1/3, Δf4 Te=1/5, Δf5 Te=1/7 and Δf6 Te=1/11. The JADE [3], SOBI [1] and FOBIUM method according to the invention are applied and the performance values αq for q=1 . . . 6 are evaluated after an averaging operation on 1000 results. For the FOBIUM method, we choose K=4 triplets of delays (τ1 k2 k3 k) where τ1 k=kTe and τ2 k3 k=0.

[0089] In the above assumptions, FIG. 5 shows the variation in α2 (performance of the second source) resulting from the JADE (b), SOBI (c) and FOBIUM (a) separators as a function of the number L of samples. The curves show firstly that the JADE and SOBI methods present difficulties in identifying the direction vector of the second source in an under-determined mixture context and that, secondly, that the FOBIUM method performs very well.

[0090]FIG. 6 gives a view, in the same context, of the variations of all the αp (1≦p≦6) values resulting from the FOBIUM method as a function of L. The curve (index p) is associated with the pth source. It can be seen that all the coefficients αp converge towards zero and that, asymptomatically, the direction vectors are perfectly identified.

[0091] Variants of the Cyclical FOBIUM Method

[0092]FIGS. 7 and 8 show two examples of variants of the method according to the invention known as the cyclical FOBIUM method.

[0093] The idea lies especially in introducing selectivity by the cyclical frequencies into the method presented here above and is aimed especially at the blind identification, with greater processing capacity, of under-determined mixtures of cyclostationary sources.

[0094] The major difference between the steps 1 to 8 explained here above and this variant is the implementation of a step for the cyclical isolation of the sources by fourth-order discrimination according to their cyclical frequencies. It is thus possible to separately identify the sources associated with a same fourth-order cyclical parameter without being disturbed by the other sources processed separately.

[0095] The two variants shown in FIGS. 7 and 8 can be implemented by reiterating the process of cyclical isolation on the “other sources” with other cyclical parameters. The process of cyclical isolation can be applied several times in a third version illustrated in FIG. 9.

[0096] Fourth-Order Cyclical Statistics

[0097] The fourth-order cyclical statistics of the observations or sensor signals used are characterized by the matrices of cyclical quadricovariance Qx ε(α,τ123) with a dimension (N2×N2) where the elements Qx ε111)[i, j, k, l], are defined by: Q x 1 ( α , τ 1 , τ 2 , τ 3 ) [ i , j , k , l ] = < Cum ( x i ( t ) , x j ( 1 - τ 1 ) * , x k ( t - τ 2 ) * , x l ( t - τ 3 ) ) exp ( - j2 πα t ) > Q x 2 ( α , τ 1 , τ 2 , τ 3 ) [ i , j , k , l ] = < Cum ( x i ( t ) , x j ( 1 - τ 1 ) * , x k ( t - τ 2 ) , x l ( t - τ 3 ) ) exp ( - j2 πα t ) > Q x 3 ( α , τ 1 , τ 2 , τ 3 ) [ i , j , k , l ] = < Cum ( x i ( t ) , x j ( 1 - τ 1 ) , x k ( t - τ 2 ) , x l ( t - τ 3 ) ) exp ( - j2 πα t ) > ( 19 )

[0098] It can be seen that Qx ε(α,τ123) is associated with the εth fourth-order moment. In stating that Qx ε(α,τ123)[i, j, k, l] is the element [N(i−1)+j, N(k−1)+l] of the matrix Qx ε(α,τ123) and assuming that the noise is Gaussian, the matrix Qx ε(α,τ123), is written as follows, in using (1) and (19): Q x 1 ( α , τ 1 , τ 2 , τ 3 ) = ( A A * ) Q s 1 ( α , τ 1 , τ 2 , τ 3 ) ( A A * ) H Q x 2 ( α , τ 1 , τ 2 , τ 3 ) = ( A A * ) Q s 2 ( α , τ 1 , τ 2 , τ 3 ) ( A A ) T Q x 3 ( α , τ 1 , τ 2 , τ 3 ) = ( A A ) Q s 3 ( α , τ 1 , τ 2 , τ 3 ) ( A A ) T ( 20 )

[0099] Where Qs ε(α,τ123) is a cyclical matrix of quadricovariance s(t) with a dimension (P2×P2), {circle over (×)} is the Kronecker product and T designates the transpose. On the assumption of statistically independent sources, the matrix Qs ε(α,τ123) is formed by at least P4−P zeros and the expression (20) is simplified as follows: Q x 1 ( α , τ 1 , τ 2 , τ 3 ) = p = 1 P c p 1 ( α , τ 1 , τ 2 , τ 3 ) ( a p a p * ) ( a p a p * ) H = A Q C s 1 ( α , τ 1 , τ 2 , τ 3 ) A Q H Q x 2 ( α , τ 1 , τ 2 , τ 3 ) = p = 1 P c p 2 ( α , τ 1 , τ 2 , τ 3 ) ( a p a p * ) ( a p a p ) T = A Q C s 2 ( α , τ 1 , τ 2 , τ 3 ) B Q T Q x 3 ( α , τ 1 , τ 2 , τ 3 ) = p = 1 P c p 3 ( α , τ 1 , τ 2 , τ 3 ) ( a p a p ) ( a p a p ) T = B Q C s 3 ( α , τ 1 , τ 2 , τ 3 ) B Q T ( 21 )

[0100] where AQ and BQ are matrices with a dimension of (N2×P) defined by AQ=[(α1{circle over (×)}α1*), . . . , (αp{circle over (×)}αp*)] and BQ=[(α1{circle over (×)}α1*), . . . , (αp{circle over (×)}αp*)], Cs ε(α, τ123) is a diagonal matrix sized (P×P) defined by Cs ε(α, τ123)=diag[c1 ε(α, τ123), . . . , c p 1 ( α , τ 1 , τ 2 , τ 3 ) = < Cum ( s p ( t ) , s p ( t - τ 1 ) * , s p ( t - τ 2 ) * , s p ( t - τ 3 ) ) exp ( - j2 π α t ) > c p 2 ( α , τ 1 , τ 2 , τ 3 ) = < Cum ( s p ( t ) , s p ( t - τ 1 ) * , s p ( t - τ 2 ) , s p ( t - τ 3 ) ) exp ( - j2 π α t ) > c p 3 ( α , τ 1 , τ 2 , τ 3 ) = < Cum ( s p ( t ) , s p ( t - τ 1 ) , s p ( t - τ 2 ) , s p ( t - τ 3 ) ) exp ( - j2 π α t ) > ( 22 )

[0101] It can be seen that the classic quadricovariance of (6) also verifies that Qx=Qx 1(0, 0, 0, 0), cp=cp 1(0, 0, 0, 0), Cs=Cs 1(0, 0, 0, 0). Recalling that Qx[i, j, k, l] is the element [N(i−1)+j, N(k−1)+l] of the matrix Qx, the following is deduced:

Qx=AQCsAQ H   (23)

[0102] In stating that Qx[i, j, k, l] is the element [N(i−1)+l, N(k−1)+j] of {tilde over (Q)}x we obtain the matrix {tilde over (Q)}x which is written as follows:

{tilde over (Q)}x=BQCsBQ H   (24)

[0103] Whitening Step

[0104] The first step of the cyclical method orthonormalizes the columns of the matrices AQ or BQ contained in the matrices Qx or {tilde over (Q)}x of the expressions (23)(24). The matrices Qx and {tilde over (Q)}x are P rank hermitian matrices and verify the following eigen-element decomposition:

Qx=ExΛxEx H et {tilde over (Q)}x={tilde over (E)}x{tilde over (Λ)}x{tilde over (Q)}x H   (25)

[0105] Where {tilde over (Λ)}x is the diagonal matrix sized (P×P) of the P non-zero values of {tilde over (Q)}x and {tilde over (E)}x is the matrix sized (N2×P) of the associated eigenvectors. For a full-rank matrix BQ, it can be shown that there is an equivalence assuming that that the kurtosis values of the sources have a same sign ε(ε=±1) and assuming that the eigenvalues of {tilde over (Λ)}x also have a same sign ε. In this context, the whitening matrix can be built according to {tilde over (T)} with a dimension of (P×N2):

{tilde over (T)}=({tilde over (Λ)}x)−1/2 {tilde over (E)} x H   (26)

[0106] where ({tilde over (Λ)}x)−1/2 is the inverse of the square root of {tilde over (Λ)}x. From the expressions (24) and (25), it is deduced that:

ε{tilde over (TQ)} x {tilde over (T)} H ={tilde over (T)}B QC s)B Q H {tilde over (T)} H =I P   (27)

[0107] This expression shows that the matrix {tilde over (T)}BQ(εCs)1/2 with a dimension (P×P) is a unitary matrix Ũ. It is then deduced from this that:

{tilde over (T)}B Q C s)−1/2   (28)

[0108] It is recalled that the whitening matrix T of Qx verifies:

TA Q =UC s)−1/2   (29)

[0109] Step of Cyclical Isolation

[0110] From the expressions (28)(29) and (21), it is deduced that: W x 1 ( α , τ 1 , τ 2 , τ 3 ) = T Q x 1 ( α , τ 1 , τ 2 , τ 3 ) T H = U ( ɛ C s ) - 1 / 2 C s 1 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C s ) - 1 / 2 U H W x 2 ( α , τ 1 , τ 2 , τ 3 ) = T Q x 2 ( α , τ 1 , τ 2 , τ 3 ) T ~ T = U ( ɛ C s ) - 1 / 2 C s 2 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C s ) - 1 / 2 U ~ T W x 2 ( α , τ 1 , τ 2 , τ 3 ) = T ~ Q x 2 ( α , τ 1 , τ 2 , τ 3 ) T ~ T = U ~ ( ɛ C s ) - 1 / 2 C s 3 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C s ) - 1 / 2 U ~ T ( 30 )

[0111] When there are P1 sources verifying ci ε(α,τ123)≠0 (1≦i≦P1), the matrix Wx ε(α,τ123) is a P1≦P ranking matrix. Thus, the unitary matrices U and Ũ with a dimension P×P may be decomposed into two sub-matrices with a dimension P×P1 and P×(P−P1) such that:

U=[U 1 U 2] and Ũ=[Ũ 1 Ũ 2]  (31)

[0112] where the matrices U1 and Ũ1 are sized P×P1 and U2 and Ũ2 are sized P×(P−P1). The matrices U1 and Ũ1 contain the singular vectors associated with the non-zero singular values of Wx ε(α,τ123). It is deduced from this that: W x 1 ( α , τ 1 , τ 2 , τ 3 ) = T Q x 1 ( α , τ 1 , τ 2 , τ 3 ) T H = U 1 ( ɛ C ~ s ) - 1 / 2 C ~ s 1 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ s ) - 1 / 2 U 1 H W x 2 ( α , τ 1 , τ 2 , τ 3 ) = T Q x 2 ( α , τ 1 , τ 2 , τ 3 ) T ~ T = U 1 ( ɛ C ~ s ) - 1 / 2 C ~ s 2 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ s ) - 1 / 2 U ~ 1 T W x 3 ( α , τ 1 , τ 2 , τ 3 ) = T ~ Q x 3 ( α , τ 1 , τ 2 , τ 3 ) T ~ T = U ~ 1 ( ɛ C ~ s ) - 1 / 2 C ~ s 3 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ s ) - 1 / 2 U ~ 1 T ( 32 )

[0113] Where the matrix {tilde over (C)}s ε(α,τ123) is a diagonal matrix sized P1×P1 formed by non-zero, diagonal elements ci ε(α,τ123) of the matrix Cs ε(α,τ123). The matrix {tilde over (C)}s={tilde over (C)}s 1 (0,0,0,0) sized P1×P1 is formed by corresponding elements ci (1≦i≦P1). Thus, after a singular value decomposition of Wx ε(α,τ123), it is possible to determine the matrices T1 and {tilde over (T)}1 from singular values associated with the non-zero singular values and T2 and {tilde over (T)}2 from singular vectors associated with the zero singular values such that:

T1=U1Π1 H, T2=U2Π2 H, {tilde over (T)}11{tilde over (Π)}1 H et {tilde over (T)}22{tilde over (Π)}2 H   (33)

[0114] where the matrices are Π1, Π2, {tilde over (Π)}1 and {tilde over (Π)}2 are unitary. From Wx ε′(α′,τ1′,τ2′,τ3′), it is possible to build a matrix {tilde over (W)}x ε′(α′,τ1′,τ2′,τ3′) depending solely on the sources of cyclical parameters (α,τ123,ε) such that ci ε(α,τ123)≠0. To do this, the following computation is made: W ~ x 1 ( α , τ 1 , τ 2 , τ 3 ) = T 1 H W x 1 ( α , τ 1 , τ 2 , τ 3 ) T 1 = Π 1 ( ɛ C ~ s ) - 1 / 2 C ~ s 1 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ s ) - 1 / 2 Π 1 H W ~ x 2 ( α , τ 1 , τ 2 , τ 3 ) = T 1 H W x 2 ( α , τ 1 , τ 2 , τ 3 ) T ~ 1 * = Π 1 ( ɛ C ~ s ) - 1 / 2 C ~ s 2 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ s ) - 1 / 2 Π ~ 1 T W ~ x 3 ( α , τ 1 , τ 2 , τ 3 ) = T ~ 1 H W x 3 ( α , τ 1 , τ 2 , τ 3 ) T ~ 1 * = Π ~ 1 ( ɛ C ~ s ) - 1 / 2 C ~ s 3 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ s ) - 1 / 2 Π ~ 1 T ( 34 )

[0115] Similarly, from Wx ε′(α′,τ1′,τ2′,τ3′) it is possible to build a matrix W ~ ~ x ɛ ( α , τ 1 , τ 2 , τ 3 )

[0116] that does not depend on the sources of cyclical parameters (α,τ123,ε) such as ci ε(α,τ123)=0: Other sources. To do this, the following computation is performed: W ~ ~ x 1 ( α , τ 1 , τ 2 , τ 3 ) = T 2 H W x 1 ( α , τ 1 , τ 2 , τ 3 ) T 2 = Π 2 ( ɛ C ~ ~ s ) - 1 / 2 < C ~ ~ s 1 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ ~ s ) - 1 / 2 Π 2 H W ~ ~ x 2 ( α , τ 1 , τ 2 , τ 3 ) = T 2 H W x 2 ( α , τ 1 , τ 2 , τ 3 ) T ~ 2 * = Π 2 ( ɛ C ~ ~ s ) - 1 / 2 C ~ ~ s 2 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ ~ s ) - 1 / 2 Π ~ 2 T W ~ ~ x 3 ( α , τ 1 , τ 2 , τ 3 ) = T ~ 2 H W x 3 ( α , τ 1 , τ 2 , τ 3 ) T ~ 2 * = Π ~ 2 ( ɛ C ~ ~ s ) - 1 / 2 C ~ ~ s 3 ( α , τ 1 , τ 2 , τ 3 ) ( ɛ C ~ ~ s ) - 1 / 2 Π ~ 2 T ( 35 )

[0117] where the matrix C ~ ~ s ɛ ( α , τ 1 , τ 2 , τ 3 )

[0118] is a diagonal matrix with dimensions (P−P1)×(P−P1) formed by diagonal elements ci ε″(α′,τ1′,τ2′,τ3′) such that the corresponding elements ci ε(α,τ123) are zero elements. The matrix C ~ ~ s

[0119] sized (P−P1)×(P−P1) is formed by the corresponding elements ci (1≦i≦P1).

[0120] In particular, in the first version of this variant, it is possible to carry out the cyclical isolation in α′=0 and ε′=1. Writing W(τ1′,τ2′,τ3′)=Wx ε′(α′,τ1′,τ2′,τ3′), {tilde over (C)}s(α′,τ1′,τ2′,τ3′)={tilde over (C)}s ε(α′,τ 1′,τ2′,τ3′) and C ~ ~ s ( τ 1 , τ 2 , τ 3 ) = C ~ ~ s ɛ ( α , τ 1 , τ 2 , τ 3 ) ,

[0121] the equations (34) and (35) become:

{tilde over (W)} x1′,τ2′,τ3′)=T 1 H W x1′,τ2′,τ3′)T 11{tilde over (C)} s)−1/2 {tilde over (C)} s1′,τ2′,τ3′)(ε{tilde over (C)} s)−1/2Π1 H   (36)

[0122] W ~ ~ x ( τ 1 , τ 2 , τ 3 ) = T 2 H W x ( τ 1 , τ 2 , τ 3 ) T 2 = Π 2 ( ɛ C ~ ~ s ) - 1 / 2 C ~ ~ s ( τ 1 , τ 2 , τ 3 ) ( ɛ C ~ ~ s ) - 1 / 2 Π 2 H ( 37 )

[0123] Identification Step

[0124] The equations (34) and (36) show that it is possible to identify the unitary matrices Π1 and {tilde over (Π)}1 associated with the sources of cyclic parameters (α,τ123) in carrying out the joint SVD (singular value decomposition) of the matrices {tilde over (W)}x ε j j1 j2 j3 j) for 1≦j≦K. Thus, to estimate the left unitary matrix, the joint diagonalizing of the matrices is performed:

{tilde over (W)} x ε j j1 j2 j3 j){tilde over (W)} x ε j j1 j2 j3 j)H for 1≦j≦K   (38)

[0125] and to estimate the right unitary matrix, the joint diagonalizing of the matrices is performed:

{tilde over (W)} x ε j j1 j2 j3 j){tilde over (W)} x ε j j1 j2 j3 j) for 1≦j≦K   (39)

[0126] To estimate the unitary matrices Π2 and {tilde over (Π)}2 associated with the sources not associated with the cyclical parameters (α,τ123,ε), the joint SVD of the matrices W ~ ~ x ɛ j ( α j , τ 1 j , τ 2 j , τ 3 j )

[0127] for 1≦j≦K is performed in jointly diagonalizing the matrices W ~ ~ x ɛ j ( α j , τ 1 j , τ 2 j , τ 3 j ) W ~ ~ x ɛ j ( α j , τ 1 j , τ 2 j , τ 3 j ) H

[0128] and then the matrices W ~ ~ x ɛ j ( α j , τ 1 j , τ 2 j , τ 3 j ) H W ~ ~ x ɛ j ( α j , τ 1 j , τ 2 j , τ 3 j ) .

[0129] Knowing Π1 and Π2, from the equation (33), the unitary matrices U1, U2 and U are deduced to the nearest permutation matrix, in performing:

U 1 =T 1Π1 , U 2 =T 2Π2 et U=[U 1 U 2]  (40)

[0130] From the equations (9) and (29), it is possible to deduce the matrix AQ to the nearest diagonal matrix and permutation matrix, such that:

T # U=[b l . . . b P ]=E xΛx 1/2 =A QC s)1/2ΛΠ  (41)

[0131] where T# is the pseudo-inverse of the matrix T. Each column, bl(1≦l≦P), of the matrix T# U is associated with a vector μq|cq|1/2q{circle over (×)}αq*), 1≦q≦P, where μq is a complex scalar value such that |μq|=1. As a consequence, in converting each column bl of the matrix T# U into a matrix B. sized (N×N) such that Bl[i, j]=bl((i−1)N+j) (1≦i, j≦N), it is deduced that:

B lq |c| 1/2αqαq H for (1≦l, q≦P)   (42)

[0132] In this context, the direction vector αq of the qth source is associated with the eigenvector of Bl associated with the greatest eigenvalue.

[0133] Recapitulation of the First Version of the Cyclical Procedure

[0134] In short, the steps of the first version of the cyclical method are summarized here below and are applied to L observations x(lTe) (1≦l≦L) of the signals received on the sensors (Te: sampling period).

[0135] Estimation

[0136] Step-1: The estimation of the matrices Qx and {tilde over (Q)}x from the L observations x(lTe). The estimation of these matrices will depend on the following assumptions:

[0137] Stationary and centered case: empirical estimator used in the reference [3].

[0138] Cyclostationary and centered case: estimator implemented in the reference [10].

[0139] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0140] Whitening

[0141] Step 2: The eigen-element decomposition of the estimates of the matrices Qx and {tilde over (Q)}x. From these operations of decomposition, the estimation of the number of sources P and use of the P main eigenvalues such that: Qx≈ExΛxEx H et {tilde over (Q)}x={tilde over (E)}x{tilde over (Λ)}x{tilde over (E)}x H where Λx and {tilde over (Λ)}x are diagonal matrices containing the P eigenvalues with the highest modulus and Ex and {tilde over (E)}x are the matrices containing the associated eigenvectors.

[0142] Step 3: The building of the whitening matrices: T=({tilde over (Λ)}x)−1/2Ex H and {tilde over (T)}=({tilde over (Λ)}x)−1/2{tilde over (E)}x H

[0143] Step 4: The selection of the cyclical parameters (α,τ123,ε) and the estimation of the matrix Qx ε(α,τ123) from the L observations x(lTe). The estimation of this matrix will depend on the following assumptions on the signals:

[0144] Stationary and centered case: empirical estimator used in the reference [3].

[0145] Cyclostationary and centered case: estimator implemented in the reference [10].

[0146] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0147] Step-5: The computation of a matrix Wx ε(α,τ123) of (30) from matrices Qx ε(α,τ123), T and {tilde over (T)}. After singular value decomposition Wx ε(α,τ123), the determining of the unitary matrices T1 and {tilde over (T)}1 associated with the non-zero singular values and T2 and {tilde over (T)}2 associated with the zero singular values.

[0148] Selection

[0149] Step-6: The selection of the K triplets of delays (τ1 k2 k3 k) where |τ1 k|+|τ2 k|+τ3 k|≠0.

[0150] Estimation

[0151] Step-7: The estimation of the K matrices Qx1 k2 k3 k) of (2). As in the step-1 this estimation will depend on the assumptions made on the signal such as:

[0152] Stationary and centered case: empirical estimator used in the reference [3].

[0153] Cyclostationary and centered case: estimator implemented in the reference [10].

[0154] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0155] Identification

[0156] Step-8: The computation of the matrices T1 Qx1 k2 k3 k) T1 H and the estimation of the unitary matrix U1 (associated with the cyclical parameters (α,τ123,ε)) in carrying out the joint diagonalizing of the K matrices: T1 Qx(α,τ1 k2 k3 k)T1 H.

[0157] Step-9: The computation of the matrices T2 Qx1 k2 k3 k) T2 H and the estimation of the unitary matrix U2 (associated with the other sources) in carrying out the joint diagonalizing of the K matrices T2 Qx1 k2 k3 k)T2 H.

[0158] Step-10: The computation of the unitary matrix U in performing: U=[U1U2]

[0159] Step-11: The computation of T#U=[bl . . . bP] and the building of the matrices Bl with a dimension (N×N) from the columns bl of T#U.

[0160] Step-12: The estimation of the signatures a, (1≦q≦P) of the P sources in applying a decomposition into elements in each matrix Bl.

[0161] Recapitulation of the Second Version of the Cyclical Procedure

[0162] The steps of the second version of the cyclical FORBIUM method are summarized here below and are applied to L observations x(lTe) (1≦l≦L) of the signals received on the sensors (Te: sampling period).

[0163] Estimation

[0164] Step-1: The estimation of the matrices Qx and {tilde over (Q)}x from the L observations x(lTe). The estimation of these matrices will depend on the following assumptions:

[0165] Stationary and centered case: empirical estimator used in the reference [3].

[0166] Cyclostationary and centered case: estimator implemented in the reference [10].

[0167] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0168] Step-2: The eigen-element decomposition of the matrices Qx and {tilde over (Q)}x. From these operations of decomposition, the estimation of the number of sources P and the use of the P main eigenvalues such that: Qx≈ExΛxEx H and {tilde over (Q)}x={tilde over (E)}x{tilde over (Λ)}x{tilde over (E)}x H where Λx and {tilde over (Λ)}x are diagonal matrices containing the P eigen values with the highest modulus and Ex and {tilde over (E)}x are the matrices containing the P associated eigen vectors.

[0169] Step-3: The building of the whitening matrices: T=(Λx)−1/2Ex H and {tilde over (T)}=({tilde over (Λ)}x)−1/2{tilde over (E)}x H

[0170] Step-4: The selection of the cyclical parameters (α, τ123,ε) and the estimation of the matrix Qx ε(α,τ123) from the L observations x((lTe). The estimation of this matrix will depend on the following assumptions on the signals:

[0171] Stationary and centered case: empirical estimator used in the reference [3].

[0172] Cyclostationary and centered case: estimator implemented in the reference [10].

[0173] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0174] Step-5: The computation of a matrix Wx ε(α,τ123) of (30) from the matrices Qx ε(α,τ123), T and {tilde over (T)}. After a singular value decomposition of Wx ε(α,τ123), the determining of the unitary matrices T1 and {tilde over (T)}1 associated with the non-zero significant values and T2 and {tilde over (T)}2 associated with the zero singular values.

[0175] Step-6: The selection of K sets of parameters (αk1 k2 k3 k).

[0176] Step-7: The estimation of the K matrices Ax εkk1 k2 k3 k) of (19). As in the step-1 this estimation will depend on the assumptions made on the signal such as:

[0177] Stationary and centered case: empirical estimator used in the reference [3].

[0178] Cyclostationary and centered case: estimator implemented in the reference [10].

[0179] Cyclostationary and non-centered case: estimator implemented in the reference [11].

[0180] Step-8: The computation of the matrices T1 Qx εkk1 k2 k3 k) T1 H and the estimation of the unitary matrix U1 or Ũ1 (associated with the cyclical parameters (α,τ123,ε)) in carrying out the joint SVD of the K matrices: T1 Qx εkk1 k2 k3 k)T1 H.

[0181] Step-9: The computation of the matrices T2 Qx εkk1 k2 k3 k)T2 H and the estimation of the unitary matrix U2 or Ũ2 (associated with the cyclical parameters (α,τ123,ε)) in carrying out the joint SVD of the K matrices: T2 Qx εkk1 k2 k3 k)T2 H.

[0182] Step-10: The computation of the unitary matrix U in performing: U=[U1U2]

[0183] Step-11: The computation of T#U=[bl . . . bP] and the building of the matrices Bl with a dimension (N×N) from the columns bl of T#U. Step-12: The estimation of the signatures aq (1≦q≦P) of the P sources in applying a decomposition into elements on each matrix Bl.

[0184] Bibliography

[0185] [1] A. Belouchrani, K. Abed—Meraim, J. F. Cardoso, E. Moulines, “A blind source separation technique using second-order statistics”, IEEE Trans. Sig. Proc., Vol.45, N°2, pp. 434-444, February 1997.

[0186] [2] J F. Cardoso, “Super-symmetric decomposition of the fourth order cumulant tensor”, ICASSP 1991.

[0187] [3] J. F. Cardoso, A. Souloumiac, “Blind beam forming for non-gaussian signals”, IEE Proceedings-F, Vol.140, N°6, pp. 362-370, December 1993.

[0188] [4] P. Chevalier, G. Benoit, A. Ferréol <<Direction finding after blind identification of sources steering vectors: The blind-maxcor and blind-MUSIC methods>>, EUSIPCO, Trieste, pp 2097-2100, 1996

[0189] [5] P. Chevalier, “Optimal separation of independent narrow-band sources: concept and performance”, Sig. Proc., Elsevier, Vol.73, pp 27-47, 1999.

[0190] [6] P. Chevalier, A. Ferréol , “On the virtual array concept for the fourth-order direction finding problem”, IEEE trans on signal processing, Vol.47, N°9, pp. 2592-2595, September 1999.

[0191] [7] P. Comon, “Independent component analysis—a new concept?”, Sig. Proc., Elsevier, Vol.36, N°3, April 1994.

[0192] [8] P. Comon “Blind channel identification and extraction of more sources than sensors”, SPIE Conf Adv Proc VIII , San Diego, July 1998.

[0193] [9] L. De Lathauwer, B. De Moor , J. Vandewalle , “ICA techniques for more sources than sensors”, Proc IEEE Processing Workshop on Higher Order Statistics, Caesarea, Israel , June 1999.

[0194] [10] A. Ferréol, P. Chevalier, “On the behavior of current second and higher order blind source separation methods for cyclostationary sources”, IEEE Trans. Sig. Proc., Vol.48, N°6, pp. 1712-1725, June 2000.

[0195] [11] A. Ferréol, P. Chevalier, L. Albera “Higher order blind separation of non zero-mean cyclostationnary sources”, (EUSPICO 2002), Toulouse, September 2002.

[0196] [12] A. Ferréol, P. Chevalier, L. Albera, “Procédé de traitement d'antennes sur des signaux cyclostationnaires potentiellement non centrés,” (Method of antenna processing on potentially non-centered cyclostationary signals), patent, May 2002.

[0197] [13] A. Ferréol, P. Chevalier “Higher order blind source separation using the cyclostationarity property of the signals”, ICASSP Munich, Vol 5, pp4061-4064, 1997

[0198] [14] A. Ferréol, P. Chevalier “Higher order blind source separation using the cyclostationarity property of the signals”, ICASSP Munich, Vol 5, pp4061-4064, 1997

[0199] [10] A. Taleb “An algorithm for the blind identification of N independent signal with 2 sensors”, 16th symposium on signal processing and its applications (ISSPA 2001), Kuala-Lumpur, August 2001.

Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US7098849 *Sep 22, 2005Aug 29, 2006Interdigital Technology CorporationBlind signal separation using array deflection
US7113129 *Sep 22, 2005Sep 26, 2006Interdigital Technology CorporationBlind signal separation using a combination of correlated and uncorrelated antenna elements
US7414579 *Sep 22, 2005Aug 19, 2008Interdigital Technology CorporationBlind signal separation using correlated antenna elements
US7548592 *Sep 30, 2004Jun 16, 2009James Stuart WightMultiple input, multiple output communications systems
US7738595Jul 2, 2004Jun 15, 2010James Stuart WightMultiple input, multiple output communications systems
US7822141Aug 31, 2006Oct 26, 2010James WightMultiple input, multiple output communications systems
US8031117 *Sep 22, 2005Oct 4, 2011Interdigital Technology CorporationBlind signal separation using polarized antenna elements
US8229018Sep 17, 2010Jul 24, 2012Zarbana Digital Fund LlcMultiple input, multiple output communications systems
US8548067 *Jan 31, 2011Oct 1, 2013Goran IvkovicSingle sensor radio scene analysis for packet based radio signals using 2nd and 4th order statistics
US8693602Nov 3, 2010Apr 8, 2014ThalesMethod and device for mono- and multi-antenna reception for Alamouti-type links
US20090093964 *Nov 17, 2006Apr 9, 2009Universite De Rennes 1Multi-dimensional parameter identification method and device: application to the location and reconstruction of deep electrical activities by means of surface observations
US20110188395 *Jan 31, 2011Aug 4, 2011Rutgers, The State University Of New JerseySingle Sensor Radio Scene Analysis for Packet Based Radio Signals using 2nd and 4th Order Statistics
US20110206792 *Oct 22, 2009Aug 25, 2011Matteo TutinoCompositions comprising vitamins
US20130301760 *May 2, 2013Nov 14, 2013Renesas Mobile CorporationMethods, apparatus and computer programs for signal interference ratio estimation
WO2006034498A2 *Sep 23, 2005Mar 30, 2006Steven J GoldbergBlind signal separation using array deflection
WO2011054857A1 *Nov 3, 2010May 12, 2011ThalesMethod and device for mono- and multi-antenna reception for alamouti-type links
Classifications
U.S. Classification375/340
International ClassificationH03H21/00, G06K9/62
Cooperative ClassificationG06K9/6243, H03H21/0012
European ClassificationG06K9/62B4N2, H03H21/00B
Legal Events
DateCodeEventDescription
Aug 23, 2011FPAYFee payment
Year of fee payment: 4
Aug 10, 2004ASAssignment
Owner name: THALES, FRANCE
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:FERREOL, ANNE;ALBERA, LAURENT;CHEVALIER, PASCAL;REEL/FRAME:015672/0479
Effective date: 20040722